Summary of the invention
The embodiment of the invention provides a kind of detection methods of pulse acoustical signal, can not be influenced, be subtracted by ambient noise
The detection false alarm rate of few acoustic detection equipment, promotes the probability and accuracy of detection.
A kind of detection method for pulse acoustical signal that first aspect of the embodiment of the present invention provides, includes the following steps:
Step 1: the pulse acoustical signal in channel each in microphone array being sampled, the sampling sound letter in each channel is obtained
Number;
Step 2: judging with the presence or absence of strong background noise in the sampling acoustical signal, if so, first believing the sampling sound
Number noise reduction process is carried out, then calculates the frame energy of the sampling acoustical signal after noise reduction process;If it is not, being adopted described in then directly calculating
The frame energy of sample acoustical signal;
Step 3: being detected according to the existence of the frame energy of the sampling acoustical signal and the sampling acoustical signal each logical
The pulse acoustical signal in road.
In conjunction with the embodiment of the present invention in a first aspect, in the first implementation of first aspect of the embodiment of the present invention,
Detecting that the pulse sound in each channel is believed according to the frame energy of the sampling acoustical signal and the existence of the sampling acoustical signal
After number, further includes: when meeting prerequisite, determine the pulse acoustical signal in each channel detected for target pulse sound letter
Number.
In conjunction with the first implementation of first aspect of the embodiment of the present invention, the second of first aspect of the embodiment of the present invention
It is described when meeting prerequisite, to determine that the pulse acoustical signal in each channel of detection is target pulse sound in kind implementation
Before signal further include: determine whether to meet prerequisite;
It is described to determine whether that meeting prerequisite includes:
Whether the relative coefficient of the pulse acoustical signal in any two channel in the pulse acoustical signal in each channel of determining detection
Greater than the first preset threshold, if not, it is determined that be unsatisfactory for prerequisite;If so, determining the pulse sound signal length in each channel
Whether the second preset threshold is greater than, if not, it is determined that be unsatisfactory for prerequisite;If so, determining the pulse acoustical signal in each channel
Signal-to-noise ratio whether be greater than third predetermined threshold value, if not, it is determined that be unsatisfactory for prerequisite, if so, determination meet preset item
Part.
In conjunction with second of implementation of first aspect of the embodiment of the present invention, in the third of first aspect of the embodiment of the present invention
It plants in implementation, after determination meets prerequisite, the target pulse acoustical signal is saved in caching, and record institute
Initial position and the end position of target pulse acoustical signal are stated, for carrying out according to the initial position and the end position
The extraction of target pulse acoustical signal;
After determination is unsatisfactory for prerequisite, each channel pulse acoustical signal that will test out is removed, and re-executes step
Rapid 1.
In conjunction with the embodiment of the present invention in a first aspect, in the 4th kind of implementation of first aspect of the embodiment of the present invention,
It whether there is strong background noise in the judgement sampling acoustical signal, if so, first carrying out noise reduction to the sampling acoustical signal
Processing includes:
Judge that the time-domain signal of the sampling acoustical signal whether there is periodically, if so, to the sampling acoustical signal
Time-domain signal carries out frequency-domain transform, if obtaining the line spectrum with harmonic relationships, carries out to the line spectrum with harmonic relationships
Inhibit.
In conjunction with the embodiment of the present invention in a first aspect, in the 4th kind of implementation of first aspect of the embodiment of the present invention,
It whether there is strong background noise in the judgement sampling acoustical signal, if so, first carrying out noise reduction to the sampling acoustical signal
Processing includes:
When the existence of the sampling acoustical signal is 0 or 1, if the frame energy of the sampling acoustical signal is greater than preset
Peak, it is determined that present sample acoustical signal is pulse acoustical signal, and present sample acoustical signal is put into the second caching, is recorded
The initial point position of present sample acoustical signal;
If the frame energy of the sampling acoustical signal is greater than preset median and is less than preset peak, it is determined that current
Sampling acoustical signal is doubtful pulse acoustical signal, and re-execute the steps 1;
If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than preset median, it is determined that current
Sampling acoustical signal is noise, and re-execute the steps 1;
When the existence of the sampling acoustical signal is 2, if the frame energy of the sampling acoustical signal is greater than in preset
Between value and be less than the peak that is set to, it is determined that present sample acoustical signal is pulse acoustical signal, and present sample acoustical signal is put
Enter in caching;
If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than the median being set to, record is current
The end point position of acoustical signal is sampled, and is calculated currently according to the initial point position of present sample acoustical signal and end point position
Sample the length of acoustical signal;
When the existence of the sampling acoustical signal is 3, postponed according to the length of calculated present sample acoustical signal
Deposit middle extraction pulse acoustical signal.
The detection device for a kind of pulse acoustical signal that second aspect of the embodiment of the present invention provides, comprising: sampling module is used for
The pulse acoustical signal in channel each in microphone array is sampled, the sampling acoustical signal in each channel is obtained;
Judgment module, for judging with the presence or absence of strong background noise in the sampling acoustical signal, if so, first being adopted to described
Sample acoustical signal carries out noise reduction process, then calculates the frame energy of the sampling acoustical signal after noise reduction process;If it is not, then directly calculating
The frame energy of the sampling acoustical signal;
Detection module, for being detected according to the frame energy of the sampling acoustical signal and the existence of the sampling acoustical signal
The pulse acoustical signal in each channel out.
In conjunction with the second aspect of the embodiment of the present invention, in the first implementation of second aspect of the embodiment of the present invention,
Described device further includes the first determining module, first determining module, for when meeting prerequisite, determines and to detect
The pulse acoustical signal in each channel is target pulse acoustical signal.
In conjunction with the first implementation of second aspect of the embodiment of the present invention, the second of second aspect of the embodiment of the present invention
In kind implementation, described device further includes the second determining module, and second determining module is used to determine whether to meet preset
Condition, second determining module are specifically used for the pulse sound in any two channel in the pulse acoustical signal for determining each channel of detection
Whether the relative coefficient of signal is greater than the first preset threshold, if not, it is determined that is unsatisfactory for prerequisite;If so, determining each
Whether the pulse sound signal length in channel is greater than the second preset threshold, if not, it is determined that be unsatisfactory for prerequisite;If so, really
Whether the signal-to-noise ratio of the pulse acoustical signal in fixed each channel is greater than third predetermined threshold value, if not, it is determined that it is unsatisfactory for prerequisite, if
It is, it is determined that meet prerequisite.
In conjunction with second of implementation of second aspect of the embodiment of the present invention, in the third of second aspect of the embodiment of the present invention
In kind of implementation, described device further includes preserving module and removes module, the preserving module be used for determine meet it is preset
After condition, the target pulse acoustical signal is saved in caching, and records the initial position of the target pulse acoustical signal
And end position, it is described with the extraction for carrying out target pulse acoustical signal according to the initial position and the end position
It removes module to be used for, after determination is unsatisfactory for prerequisite, each channel pulse acoustical signal that will test out is removed.
In conjunction with the second aspect of the embodiment of the present invention, in the 4th kind of implementation of second aspect of the embodiment of the present invention,
The judgment module is specifically used for judging that the time-domain signal of the sampling acoustical signal whether there is periodically, if so, to described
The time-domain signal for sampling acoustical signal carries out frequency-domain transform, if obtaining the line spectrum with harmonic relationships, to described there is harmonic wave to close
The line spectrum of system is inhibited.
In conjunction with the second aspect of the embodiment of the present invention, in the 5th kind of implementation of second aspect of the embodiment of the present invention,
The detection module is specifically used for:
When the existence of the sampling acoustical signal is 0 or 1, if the frame energy of the sampling acoustical signal is greater than preset
Peak, it is determined that present sample acoustical signal is pulse acoustical signal, and present sample acoustical signal is put into the second caching, is recorded
The initial point position of present sample acoustical signal;
If the frame energy of the sampling acoustical signal is greater than preset median and is less than preset peak, it is determined that current
Sampling acoustical signal is doubtful pulse acoustical signal, and re-execute the steps 1;
If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than preset median, it is determined that current
Sampling acoustical signal is noise, and re-execute the steps 1;
When the existence of the sampling acoustical signal is 2, if the frame energy of the sampling acoustical signal is greater than in preset
Between value and be less than the peak that is set to, it is determined that present sample acoustical signal is pulse acoustical signal, and present sample acoustical signal is put
Enter in caching;
If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than the median being set to, record is current
The end point position of acoustical signal is sampled, and is calculated currently according to the initial point position of present sample acoustical signal and end point position
Sample the length of acoustical signal;
When the existence of the sampling acoustical signal is 3, postponed according to the length of calculated present sample acoustical signal
Deposit middle extraction pulse acoustical signal.
The method of a kind of pulse acoustical signal processing provided in an embodiment of the present invention, including to channel each in microphone array
Pulse acoustical signal is sampled, and the sampling acoustical signal in each channel is obtained, and judges to make an uproar in the sampling acoustical signal with the presence or absence of strong background
Sound if so, first carrying out noise reduction process to the sampling acoustical signal, then calculates the frame energy of the sampling acoustical signal after noise reduction process
Amount;If it is not, the frame energy of the sampling acoustical signal is then directly calculated, according to the frame energy of the sampling acoustical signal and the sampling acoustical signal
Existence detect the pulse acoustical signal in each channel, compared with the existing technology, the embodiment of the present invention is in detection pulse sound letter
Number when can carry out noise reduction process, influence of the noise in environment to testing result can be effectively excluded, to promote detection
The probability and accuracy of pulse acoustical signal.
Description and claims of this specification and the (if present)s such as term " first " in above-mentioned attached drawing, " second "
It is to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that the number used in this way
According to being interchangeable under appropriate circumstances, so as to the embodiments described herein can in addition to the content for illustrating or describing herein with
Outer sequence is implemented.In addition, term " includes " and " having " and their any deformation, it is intended that cover non-exclusive packet
Contain, for example, what the process, method, system, product or equipment for containing a series of steps or units were not necessarily limited to be clearly listed
Those step or units, but may include be not clearly listed or it is intrinsic for these process, methods, product or equipment
Other step or units.
The first embodiment of pulse acoustical signal processing method in the embodiment of the present invention is introduced first below, the pulse acoustical signal
Detection method can be applied in military field, for detecting the pulse acoustical signal of shot, cannon sound, explosive sound.
As shown in Figure 1, the detection method of the pulse acoustical signal includes:
Step 101 samples the pulse acoustical signal in channel each in microphone array, obtains the sampling sound letter in each channel
Number.
In the present embodiment, the pickup of signal is completed in the form of more microphones form array first, obtains each channel
Pulse acoustical signal.Array structure composition includes 4 microphone array element Isosorbide-5-Nitrae root horizontal array bars 2, referring specifically to Fig. 2.Then right
Each microphone channel is sampled by frame, and frame length is rule of thumb set, and has just been started to acquire several frame data and is not processed, only into
Row accumulation, sample frequency can be preset.
It should be noted that in other embodiments, also needing to carry out parameter initialization, parameter before executing step 1
Including threshold value PWMTH, frame energy in signal existence ST, frame energy threshold PWTH, frame energy Low threshold PWLTH, frame energy
High threshold PWHTH, array signal correlation coefficient threshold CFTH, frequency domain snr threshold FSNRTH, maximum impulse wave persistence length
Threshold value MAXLS, signal existence ST are initially 0, and frame energy threshold PWTH is initially empirically determined value or last time
Value before equipment shutdown, PWLTH=1.5 × PWTH, PWMTH=3 × PWTH, PWHTH=6 × PWTH, correlation coefficient threshold
CFTH=0.3, frequency domain snr threshold FSNRTH=7dB, maximum impulse wave persistence length threshold value MAXLS=0.5 × fs, fs are
Sample rate.
Step 102 judges with the presence or absence of strong background noise in the sampling acoustical signal, if so, 103 are thened follow the steps, if
It is no, then follow the steps 104;
Step 103 first carries out noise reduction process to the sampling acoustical signal, then calculates the letter of the sampling sound after noise reduction process
Number frame energy;
Step 104, the frame energy for directly calculating the sampling acoustical signal;
In the present embodiment, by the analysis to existing detection technique, unite when signal detection using signal time domain
Metering or frequency domain statistic, so must carry out certain processing to sampled signal extracts the amount that can be used for detecting, this hair
It is bright using the energy for seeking certain frame length time-domain signal, calculation formula is as follows:
In formula (2), N is frame length, x (i), i=1 ... N sequence of time-domain samples, | | absolute value is sought in expression.
Strong background noise in the present embodiment includes the noise of oil machine noise, Vehicle Engine Noise this kind, acoustic detection
Equipment needs to be placed on when in use in many cases beside power generation oil machine, or is placed on the roadside of running car, this can all make
At there are strong background noises in environment near acoustic detection equipment, when having pulsed sound sending, microphone array is arrived in addition to that can pick up
Pulsed sound can also pick up strong background noise signal, will lead to useful pulse sound and be submerged in strong noise background completely, such as
For fruit without processing, manually vision can not all find pulse acoustical signal, be difficult the arteries and veins being tested with by automatic detection algorithm
Acoustical signal is rushed, so strong background noise inhibition must be carried out.
Step 105 detects respectively according to the frame energy of the sampling acoustical signal and the existence of the sampling acoustical signal
The pulse acoustical signal in channel.
The second embodiment of pulse sound signal detecting method provided by the present invention is described below, as shown in figure 3,
The detection method of the pulse acoustical signal includes:
Step 201 samples the pulse acoustical signal in channel each in microphone array, obtains the sampling sound letter in each channel
Number.
The method that the executive mode of step 201 can refer to step 101 in the present embodiment, does not limit specifically specifically herein
It is fixed.
Step 202 judges that the time-domain signal of the sampling acoustical signal whether there is periodically, if so, thening follow the steps
203。
In the present embodiment, sampling in acoustical signal may include strong background noise, and strong background noise includes oil machine noise, vehicle
The noise of engine noise this kind can pass through since the time-domain signal of strong background noise has the characteristics that periodic
Whether the time-domain signal of judgement sampling acoustical signal whether there is periodically, so that it is determined that including that strong background is made an uproar in sampling acoustical signal
Sound.Such as when oil machine the operation is stable, it is as shown in Figure 4 that sampling obtains time-domain signal, it can be seen that has the apparent period
Property.
Step 203 carries out frequency-domain transform to the time-domain signal of the sampling acoustical signal, if obtaining the line with harmonic relationships
Spectrum, then inhibit the line spectrum with harmonic relationships.
It in the present embodiment, is illustrated so that strong background noise is oil machine noise as an example, includes when determining in sampling acoustical signal
When oil machine noise, it is as shown in Figure 5 that frequency-region signal is obtained to the time-domain signal progress Fourier transformation of oil machine noise, it can be seen that tool
There is apparent line spectrum feature, and all line spectral frequencies are all based on the same fundamental frequency, the fundamental frequency of line spectrum is 25Hz in Fig. 5,
Acoustic signals can change after air borne, and the variation of harmonic relationships is no more than 1Hz, this kinds of noise main energetic all collects
In on the line spectrum with harmonic relationships, as long as so eliminating most of noise after inhibiting to line spectrum.
This line spectrum noise with harmonic wave is inhibited to find out harmonic wave collection firstly the need of in frequency domain, it is true to each harmonic
Determine narrow band bandwidth, in entire spectral range, retain the corresponding amplitude of each Frequency point except each narrow band bandwidth, in each narrow band bandwidth
The corresponding amplitude of each Frequency point using batten difference method determine, finally carry out frequency domain inverse transformation obtain time domain waveform, i.e.,
For the signal after noise suppressed.
When line-spectrum detection, signal spectrum is normalized first, with the following method search spectrum peak:
1. detects the maximum local_max in frequency spectrum spectrum;
2. respectively finds minimum a local_min1 and localmin2 in maximum Dynamic Neighborhood, if 2 ×
Local_max/ (local_min1+localmin2) is greater than some threshold value, it is judged that spectral peak is significantly, to can choose.
3. carries out the detection of harmonic wave collection according to following formula:
|frequency(k)-fm×imk|≤ε (2)
Wherein frequency (k) indicates the corresponding frequency of spectral peak, fmThe fundamental frequency assumed that, imkCorresponding overtone order, ε are
The threshold value of setting.If detecting some frequencies and a certain specific frequency having apparent harmonic relationships, then it is assumed that detect one it is humorous
Wave collection, while finding corresponding fundamental frequency.
When harmonic relationships are not present in signal, the biggish line spectrum of amplitude is inhibited.Fig. 6, which is shown, have been collected by force
The pulse sound time-domain signal of ambient noise, as can be seen from the figure pulse sound Signal-to-Noise is lower, is unfavorable for target detection, figure
7 show the time-domain signal after noise suppressed, and the biggish promotion that as can be seen from the figure pulse sound Signal-to-Noise obtains has
Conducive to the work of subsequent target detection.
Step 204, after inhibiting to the line spectrum with harmonic relationships, calculate it is described sampling acoustical signal frame energy
Amount;
Step 205 detects respectively according to the frame energy of the sampling acoustical signal and the existence of the sampling acoustical signal
The pulse acoustical signal in channel.
In the present embodiment, step 204 includes:
S2051, when it is described sampling acoustical signal existence be 0 or 1 when, if it is described sampling acoustical signal frame energy be greater than
Preset peak, it is determined that present sample acoustical signal is pulse acoustical signal, and present sample acoustical signal is put into the second caching
In, record the initial point position of present sample acoustical signal;If it is described sampling acoustical signal frame energy be greater than preset median and
Less than preset peak, it is determined that present sample acoustical signal is doubtful pulse acoustical signal, and re-execute the steps 1;If described
The frame energy for sampling acoustical signal is greater than preset minimum and is less than preset median, it is determined that present sample acoustical signal is to make an uproar
Sound, and it re-execute the steps 1;
S2052, when the existence of the sampling acoustical signal is 2, if the frame energy of the sampling acoustical signal be greater than it is pre-
If median and be less than the peak that is set to, it is determined that present sample acoustical signal is pulse acoustical signal, and by present sample sound
Signal is put into caching;If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than the median being set to,
The end point position of present sample acoustical signal is recorded, and is counted according to the initial point position of present sample acoustical signal and end point position
Calculate the length of present sample acoustical signal;
S2053, when it is described sampling acoustical signal existence be 3 when, according to the length of calculated present sample acoustical signal
Degree extracts pulse acoustical signal from caching.
It should be noted that the pulse acoustical signal that the sound sources such as cannon generate is not ideal pulse sound, in the time domain can
Continue regular hour length, experienced from no signal, signal occurs, signal is lasting, signal terminates such signal there are shapes
State course, corresponding short time frame energy can also undergo the signal energy state course as low energy, high-energy, the low energy,
So binding signal existence and signal energy state course propose to sample the detection of double states in pulse acoustical signal processing
Method, signal energy state are determined that signal energy state is controlled by high, medium and low frame energy threshold (PWLTH, PWMTH, PWHTH)
The evolution of signal existence.
What microphone received before pulse acoustical signal occurs is all that ambient noise signal frame energy is lower, so setting low frame
Energy threshold indicates environmental noise power, indicates that doubtful signal occurs when frame energy is greater than low frame energy threshold, by signal
Duration impulse acoustical signal necessarily will appear bigger frame energy, so when frame energy be greater than high frame energy threshold when indicate determine
Signal occurs, and after determining that signal occurs, continuing frame energy greater than middle frame energy threshold indicates that persistent signal occurs, when frame energy is small
Indicate that signal terminates when middle frame energy threshold, detection finishes.
A certain size caching buff1 is set, storage environment noise data is used for, takes out ambient noise when buff1 is full
Data calculate ambient noise average energy, as frame energy threshold PWTH.Since the influence pulse sound of region, landform is believed
Different degrees of reverberation can be generated when number being transferred to microphone receiving area, causes smear longer, so frame energy in setting
Thresholding PWMTH terminates to detect for signal, and PWMTH is more larger than for detecting low frame energy threshold PWLMTH when signal occurs,
The hangover of signal can be reduced to a certain extent in this way.Biggish caching buff2 is set, for storing the pulse being confirmed
Acoustical signal, buff2 will not be filled with after pulse acoustical signal processing, so initial point position X1 and end point by recording
Signal is taken out in X2 determination in position from buff2.
Signal extraction finishes and signal existence ST, initial point position X1, end point position after further confirming that
X2, which is reset, to be resetted.
It is the flow diagram of the 3rd embodiment of pulse sound detection method of the present invention referring to Fig. 8, Fig. 8.
Based on the first embodiment of pulse sound detection method, this method further include:
Step 106 when meeting prerequisite, determines that the pulse acoustical signal in each channel detected is that target pulse sound is believed
Number.
Wherein it is determined that the step of meeting prerequisite, includes:
1061, the relative coefficient of the pulse acoustical signal in any two channel in the pulse acoustical signal in each channel of detection is determined
Whether the first preset threshold is greater than, if not, it is determined that be unsatisfactory for prerequisite;If so, determining the pulse acoustical signal in each channel
Whether length is greater than the second preset threshold, if not, it is determined that is unsatisfactory for prerequisite;If so, determining the pulse sound in each channel
Whether the signal-to-noise ratio of signal is greater than third predetermined threshold value, if not, it is determined that is unsatisfactory for prerequisite, meets in advance if so, determining
Set condition.
It should be noted that by signal detection and the pulse sound that extracts sometimes and target pulse acoustical signal, herein
Target pulse acoustical signal includes cannon sound, shot or explosive sound, after noise signal caused by natural wind is received by microphone
Sometimes and a kind of pulse sound, so needing to further confirm that the signal of extraction.The present embodiment uses multichannel microphone array
First signal correlation method differentiates whether the pulse sound detected is wind noise.It is if each channel microphone array element is collected
The pulse sound of the same target, then should have very strong correlation;And wind noise is possible to cause on the microphone of part
Pulse sound, at this moment the correlation of microphone array element two-by-two will not be all very high, if relative coefficient is less than the first preset threshold
The case where CFTH, is unsatisfactory for prerequisite, and regards as noise.
Wherein steps are as follows for correlation analysis:
A. power spectrumanalysis is carried out to each channel time domain waveform of extraction, obtains each channel power spectrum waveform, calculation formula
It is as follows:
In formula, x (n) is sequence of time-domain samples, PXIt (w) is the power spectrum acquired, c (n) is window function.
B. channel power spectral correlative coefficient two-by-two is calculated, calculation formula is as follows:
In formula, variance is sought in cov () expression,Indicate each channel power spectrum.
C. the smallest related coefficient is extracted, judges whether to be less than threshold value CFTH, determines whether the signal extracted is wind noise.
Above-mentioned correlation analysis is carried out in frequency domain, is because each array element receives when acoustic array size is larger
The target acoustic signal time delay arrived is larger, directly asks related coefficient also smaller in time domain, is illustrated in figure 9 wherein 3 array elements and receives
It is corresponding power spectrum waveform that the same pulse sound arrived, which is 0.07, Figure 10 by calculating time domain least correlativing coefficient, by meter
Calculating frequency domain least correlativing coefficient is 0.98, so frequency domain correlation coefficient is demonstrated by array received to signal truth.
Optionally, if relative coefficient is greater than the first preset threshold CFTH, continue the pulse sound letter for determining each channel
Whether number length is greater than the second preset threshold, if the length of pulse acoustical signal is greater than the second preset threshold, it is believed that pulse sound letter
There are reverberation in number, are unsatisfactory for prerequisite, need to carry out reverberation elimination at this time.
The method that reverberation is eliminated specifically: regard reverberation space as a reverberation system, the pulse sound that sound source issues is done
For system input, by reflecting to form reverberation sound as the output of system, then the reverberation system model can be stated are as follows:
In formula, h (t) indicates the impulse response of reverberation system, and n (t) indicates that reverberation space environment noise, s (t) indicate to reach
The sound source pulse sound in reverberation space, x (t) indicate the reverberation sound of reverberation system output.
It can be seen that, reverberation system impulse response and sound-source signal are convolution relation in the time domain, to handle from above formula
Reflected sound is removed from sound-source signal must unlock this convolution relation, and cepstrum technology can linearly be divided convolution signal
From at additive signal signal is reflected in high frequency, so carrying out low pass filtered in cepstrum domain in cepstrum domain sound-source signal generally in low frequency
Dereverberation signal can be obtained in inverse transformation back signal after wave, and steps are as follows for calculating:
1. carries out Complex Cepstrum Transform to the waveform of Detection and Extraction, transformation for mula is as follows:
If the pulse sound sequence of time-domain samples of Detection and Extraction be x (k), k=1,2 ..., N, sequence length N, to sequence into
Row discrete Fourier transform:
Ask the amplitude A (k) and phase of X (k)
In formula, real () indicates to seek the real part of complex signal, and imag () indicates to seek the imaginary part of complex signal, to phaseIncome linearization process obtainsThen the cepstrum of x (k) may be expressed as:
2. carries out low-pass filtering to CX (k) and obtains signal LCX (k);
3. carries out the signal after cepstrum inverse transformation obtains dereverberation to LCX (k)Inverse transformation process and cepstrum become
Process is changed to be omitted here on the contrary.
Take reverberation example as shown in figure 11 for actual pulse sound, as can be seen from the figure the former reverb signal duration compared with
It is long, it is able to restore original shorter pulse acoustical signal after dereverberation.
Optionally, however, it is determined that the pulse sound signal length in each channel is not more than the second preset threshold, that is, thinks that pulse sound is believed
Reverberation is not present in number, further determines that whether the signal-to-noise ratio of the pulse acoustical signal in each channel is greater than third predetermined threshold value, if really
The signal-to-noise ratio of the pulse acoustical signal in fixed each channel is greater than third predetermined threshold value, then it is assumed that main energetic concentrates on limited in a frequency domain
In the frequency range of bandwidth, that is, meet prerequisite, may thereby determine that the pulse acoustical signal detected is target acoustic signal, the target
Acoustical signal includes cannon sound, shot or explosive sound.
Wherein, judged whether main energetic concentrates on limited band in a frequency domain using frequency domain signal-to-noise ratio in the present embodiment
In wide frequency range, it is as follows to define frequency domain signal-to-noise ratio:
In formula, X (k) indicates that the discrete spectrum of the pulse sound extracted, N indicate discrete spectrum length, and Lf indicates pulse sound
Lower-frequency limit, Hf indicate the upper frequency limit of pulse sound.When condition FSNR > FSNRTH meets, final confirmation detects target arteries and veins
Rush acoustical signal.
Step 107 meets prerequisite determining, and determines that the pulse acoustical signal in each channel detected is target pulse
After acoustical signal, the target pulse acoustical signal is saved in caching, and records the start bit of the target pulse acoustical signal
It sets and end position, with the extraction for carrying out target pulse acoustical signal according to the initial position and the end position.
Step 108, after determining and being unsatisfactory for prerequisite, each channel pulse acoustical signal for will test out is removed, and is laid equal stress on
It is new to execute step 101.
The method of a kind of pulse acoustical signal processing provided in an embodiment of the present invention, including to channel each in microphone array
Pulse acoustical signal is sampled, and the sampling acoustical signal in each channel is obtained, and judges to make an uproar in the sampling acoustical signal with the presence or absence of strong background
Sound if so, first carrying out noise reduction process to the sampling acoustical signal, then calculates the frame energy of the sampling acoustical signal after noise reduction process
Amount;If it is not, the frame energy of the sampling acoustical signal is then directly calculated, according to the frame energy of the sampling acoustical signal and the sampling acoustical signal
Existence detect the pulse acoustical signal in each channel, compared with the existing technology, the embodiment of the present invention is in detection pulse sound letter
Number when can carry out noise reduction process, influence of the noise in environment to testing result can be effectively excluded, to promote detection
The probability and accuracy of pulse acoustical signal.
The embodiment of pulse acoustical signal processing device in the embodiment of the present invention is described below.
Figure 12 is please referred to, the first embodiment of the detection device of pulse acoustical signal includes: in the embodiment of the present invention
Sampling module 201 samples for the pulse acoustical signal to channel each in microphone array, obtains each channel
Sample acoustical signal;
Judgment module 202, for judging with the presence or absence of strong background noise in the sampling acoustical signal, if so, first to institute
It states sampling acoustical signal and carries out noise reduction process, then calculate the frame energy of the sampling acoustical signal after noise reduction process;If it is not, then directly
Calculate the frame energy of the sampling acoustical signal;
Optionally, judgment module 202 is specifically used for whether judging the time-domain signal for sampling acoustical signal in the present embodiment
There are periodicity, if so, frequency-domain transform is carried out to the time-domain signal of the sampling acoustical signal, if obtaining having harmonic relationships
Line spectrum then inhibits the line spectrum with harmonic relationships.
Detection module 203, for according to the frame energy of the sampling acoustical signal and the existence of the sampling acoustical signal
Detect the pulse acoustical signal in each channel.
Optionally, in the present embodiment, detection module 203 is specifically used for when the existence of the sampling acoustical signal being 0 or 1
When, if the frame energy of the sampling acoustical signal is greater than preset peak, it is determined that present sample acoustical signal is pulse acoustical signal,
And present sample acoustical signal is put into the second caching, record the initial point position of present sample acoustical signal;
If the frame energy of the sampling acoustical signal is greater than preset median and is less than preset peak, it is determined that current
Sampling acoustical signal is doubtful pulse acoustical signal, and re-execute the steps 1;
If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than preset median, it is determined that current
Sampling acoustical signal is noise, and re-execute the steps 1;
When the existence of the sampling acoustical signal is 2, if the frame energy of the sampling acoustical signal is greater than in preset
Between value and be less than the peak that is set to, it is determined that present sample acoustical signal is pulse acoustical signal, and present sample acoustical signal is put
Enter in caching;
If the frame energy of the sampling acoustical signal is greater than preset minimum and is less than the median being set to, record is current
The end point position of acoustical signal is sampled, and is calculated currently according to the initial point position of present sample acoustical signal and end point position
Sample the length of acoustical signal;
When the existence of the sampling acoustical signal is 3, postponed according to the length of calculated present sample acoustical signal
Deposit middle extraction pulse acoustical signal.
3, Figure 13 is the structural schematic diagram of the second embodiment of pulse sound detection device of the present invention referring to Fig.1.
Based on the first embodiment of pulse sound signal supervisory instrument, the device further include:
First determining module 204 determines that the pulse acoustical signal in each channel detected is for when meeting prerequisite
Target pulse acoustical signal.
Second determining module 205, second determining module are used to determine whether to meet prerequisite, second determination
Module is specifically used for the relative coefficient of the pulse acoustical signal in any two channel in the pulse acoustical signal for determining each channel of detection
Whether the first preset threshold is greater than, if not, it is determined that be unsatisfactory for prerequisite;If so, determining the pulse acoustical signal in each channel
Whether length is greater than the second preset threshold, if not, it is determined that is unsatisfactory for prerequisite;If so, determining the pulse sound in each channel
Whether the signal-to-noise ratio of signal is greater than third predetermined threshold value, if not, it is determined that is unsatisfactory for prerequisite, meets in advance if so, determining
Set condition.
Preserving module 206, the preserving module are used for after determination meets prerequisite, and the target pulse sound is believed
It number is saved in caching, and records initial position and the end position of the target pulse acoustical signal, for being risen according to described
The extraction of beginning position and end position progress target pulse acoustical signal;
Module 207 is removed, for after determination is unsatisfactory for prerequisite, each channel pulse acoustical signal that will test out to be clear
It removes.
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although referring to before
Stating embodiment, invention is explained in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features;And these
It modifies or replaces, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.